This technology extracts the skeleton data of the operator through videos taken on a manufacturing site. In addition, it establishes the standard operation action database that can determine the safety of the worker’s operation. For example, according to the correctness of the process of loading, unloading, spraying, picking, and transferring in the operating press of the stamping machine, the system can establish a safety judgment model for the operation of highly dangerous machinery and tools.

Image recognition technology is used to analyze the videos of the on-site punching machine operators through the smoothing of the images, labelling and de-noising skeleton information. In addition, three processes are utilized: (1) human body joint calibration and coordination capture during the operation under time sequence map establishment; (2) use of image processing technologies such as region selection, image smoothing and image de-noising to improve computing efficiency and image recognizability; (3) three-stage process of action standardization establishment is performed for each individual operation. Actions in videos are transformed into data and visualized to conduct machine learning training and operating action recognition models. Once operators have unsafe actions during operations, the recognition system can issue warnings in a timely manner to remind operators and management personnel to pay attention to avoid problems during operations.

線上展網址：
https://tievirtual.twtm.com.tw/iframe/2ee33626-ee87-4d0f-9881-f273cff46338?group=3a1a21ad-59dd-4f93-9cbd-9b4a13294f3c&lang=en